What Can Comparative Research Tell Us About Future European Referendums? Lawrence LeDuc

Transcription

1 What Can Comparative Research Tell Us About Future European Referendums? Lawrence LeDuc Department of Political Science, University of Toronto 100 St. George St., Toronto CANADA M5S 3G3 Tony Blair s sudden U-turn last April on the issue of holding a referendum on a new EU constitution means that British voters can anticipate the prospect of at least two future referendums on European matters the vote on a prospective new EU constitution taking its place beside the long promised referendum on the Euro. Together with Britain, a number of other European countries are also likely to hold referendums on the new constitution at some point. Denmark and Ireland will be required to do so by their own constitutions, and The Netherlands, Luxembourg, Portugal and Spain are likely to hold such votes. France may as well, depending on the route chosen for approval of the new constitution. Prospects for referendums among the new member states are more uncertain, but votes in Poland and the Czech Republic seem more than probable. Many others may follow, since it becomes more difficult to make a case for preventing one s own citizens from having their say in a referendum when others are doing so. Thus, while the prospects of Britain joining the Eurozone, or Europe celebrating a new constitutional order, seem uncertain at the present time, the outlook for students of referendums appears bright indeed. In a country where referendums used to be rare political events, they now represent something of a growth industry for students of voting behaviour. Outside of Switzerland, referendums used to be rare events in much of the rest of Europe as well. But the political calculations in a number of countries now tend to favour dealing with important European issues in this way, even when there is no legal or constitutional requirement that they do so. One compelling reason is that political parties are often divided on European issues, and holding a referendum becomes one way of managing these internal party divisions. Morel (1993, 2001) notes that this is one of the more common reasons why governing parties opt for referendums on certain types of issues. They can also sometimes be a convenient way of extricating a party from previous political commitments that have become inconvenient. Harold Wilson s decision to hold a referendum on EC membership shortly after returning to power in 1974 provides a good example of this strategy. Referendums can also be a way of legitimizing a political decision that might otherwise be a source of continuing conflict. The people have spoken provides a compelling (albeit not always final) answer to certain types of difficult political questions. Then, there is the matter of precedent. Once a referendum has taken place on an important European issue, it becomes more difficult to argue against holding one on the next important issue that comes along. Labour s 1975 referendum created such a precedent in Britain, as did François Mitterrand s decision to hold a referendum on Maastricht. Once the first steps down this path are taken, referendums tend to lead to more referendums! Over time, they can become a kind of quasi-constitutional norm for dealing with certain categories of issues, e.g. institutional changes, international treaties, or major European issues, even in the absence of any formal legal requirement that such issues be handled in this way.

2 Blair s decision to abandon his previous position of avoiding a referendum on the constitution while promising (but postponing) one on the Euro probably fits several of these categories. Certainly, domestic politics rather than European concerns likely had more to do with the decision, painful though it must have been for a prime minister facing so many other challenges. Yet, even though governments are sometimes pushed into referendums that they would prefer to avoid, they seldom call referendums that they expect to lose. Timing therefore becomes all important. Like separatist Quebec governments in recent years, timing (or postponing) the referendum to coincide with what are perceived to be winning conditions is likely to be the default strategy, particularly with regard to the Euro. But even with the aid of sophisticated polling, winning conditions can be difficult to gauge. The governments of both Denmark (2000) and Sweden (2003) held their referendums on the Euro at a time that they felt they could be won. But both lost by substantial margins. At present, the chances of a British government actually winning a referendum on either the Euro or on the EU constitution would not appear to be very high. But the comparative research on referendums teaches us that campaigns are important, and that there is often considerably more volatility in a referendum campaign than in a parliamentary election (LeDuc 2002, 2003). It also shows us that context is all important. Public opinion polls can give us readings on mass opinion on an issue, but they can tell us little about the way in which that issue will be framed in a referendum campaign, or about the manner in which such a campaign will unfold. Few would have believed at the time that Felipe Gonzàlez would be able to win the 1986 referendum on NATO in Spain. Fewer still would have thought that Mitterrand would come so close to losing a referendum on Maastricht. In this paper, I will examine a number of contextual factors that are often crucial in determining the outcome of a referendum, paying particular attention to evidence from those cases that can help us to better understand the characteristics of referendum campaigns and the factors which tend to influence voting behaviour in such settings. Particularly useful in this regard are the Danish and Swedish referendums on the Euro, the 1992 referendums on Maastricht, the 1994 referendums in Austria and the Nordic countries on EU membership, the 1992 Canadian constitutional referendum, the 1999 Australian referendum on the monarchy, and the two Quebec referendums. One s first impression is that the prospects for winning a referendum in Britain either on the Euro or on the EU constitution would not appear to be very high. However, questions from the 2001 BES show that there is a large pool of voters to whom an appeal might be directed on both issues. First, with regard to Europe, opinion in Britain is still generally more favourable than unfavourable towards the EU, although it has been declining in recent years (Clarke et al, 2004). A majority of British voters continue to approve of membership in the EU, while slightly more than a third disapprove (table 1). If a YES vote on the constitution is thought of as simply affirming Britain s place in Europe, it may not be as difficult as it appears to win the support of those voters who are positively predisposed towards Europe. There is no assurance of course that the issue will be framed in this way. The Euro would appear to present the larger challenge, perhaps because it has already been the subject of considerable public debate in Britain. Only a small number of 2001 BES respondents (15%) position themselves as definitely in favour of the Euro, while nearly 40% say that they are definitely against it. But the largest number, just 2

3 under 44%, appears to accept the argument that it is the timing of British entry into the Eurozone that is important. The support of such voters might be won in a referendum if the issue can be successfully framed in terms of the timing of British entry being right. But this would not be easy to accomplish in the face of an electorate which appears to contain such a large number of voters already predisposed against the Euro. Nevertheless, there are some parallels to the 1986 setting in Spain, in which Gonzàlez faced an electorate seemingly strongly opposed in principle to the prospect of Spanish membership in NATO. With regard to the Euro at least, Blair will face no less of a challenge. Table 1. Attitudes toward the European Union and the Euro* a. Membership in the EU b. Joining the Eurozone % % Strongly approve 10.0 Definitely join 14.9 approve 49.3 Wait and see 43.5 Disapprove 27.4 Definitely stay out 39.9 Strongly disapprove 8.4 DK, other 1.7 DK, other * 2001 British Election Study, campaign wave (Gallup). N=4810. Theoretical issues Zaller s (1992) model of opinion formation is particularly well suited to the study of public opinion and voting behavior in referendums. As he argues, any process of opinion formation proceeds from an interaction of information and predisposition. The extent to which basic values and beliefs are linked to a referendum issue in public debate provides a key starting point for any theoretical understanding of referendum voting. Bützer and Marquis (2002) applied such a model to the wide range of issues which regularly confront voters in Swiss referendums. A similar dynamic might be expected in British referendums on the Euro or an EU constitution. For some voters, opinions on European integration might indeed reflect strongly held beliefs about the nation or a fundamental sense of political identity. For others however, attitudes might be less the product of beliefs developed over a period of time than a more transient opinion based on the persuasive arguments of an advertising campaign, apprehensions about the state of the economy, or judgments about the relative credibility of those delivering the campaign message. When strongly held predispositions are merely reinforced by the rhetoric of a 3

4 campaign, referendums begin to take on some of the characteristics of elections, in which factors such as party identification or ideological orientation typically play a more crucial role (Tonsgaard, 1992). But when parties are internally divided, ideological alignments are unclear, or an issue is a new and unfamiliar one, voters might be expected to draw more of their information from the campaign discourse. Under these circumstances, the outcome of the contest becomes more uncertain. Figure 1. Elements leading toward stability or volatility in referendum voting STABILITY VOLATILITY Social cleavages Campaign events Groups Government Ideology actions Media Core beliefs Political parties Leaders State of the economy Factors such as party identification, the linkage of the referendum issue to particular groups, or its identification with established political actors, give us operational examples of Zaller s predispositions. Figure 1 provides a conceptual map, on which a number of the relevant variables are arranged to fit the context of referendum voting as it might vary from issue to issue or case to case. I have argued elsewhere (LeDuc 2002b, 2003) that the closer a particular referendum comes to involving elements at the left hand side of the diagram, the more its outcome is likely to be driven by predispositions and the more limited (or reinforcing) the effects of the campaign. As one moves towards the right hand side of the diagram, the potential for movement over the course of the campaign increases and the outcome becomes progressively more uncertain. In those instances where the issue(s) of the referendum are entirely new to the voter, the learning process of the campaign thus becomes highly critical to the determination of the outcome. Bowler & Donovan (1998, 2002) note that voters draw upon a variety of different sources in forming opinions about the sometimes complex and confusing initiatives which appear on many U.S. state ballots. Among the most frequently mentioned sources of such information are campaign pamphlets, television advertising, and direct mailings from various campaign organizations. Voters in such situations take cues from these and other sources, and can often find short-cuts that enable them to cut through large amounts of sometimes 4

5 conflicting information (Bowler and Donovan 1998, 2002; Lupia and McCubbins,1998). On an issue such as the constitution, such short-cuts may be of critical importance, since relatively few voters will immerse themselves in the murky (and often contested) details of a legal document. Constitutional referendums in Australia, for example, have often been determined more by partisahship or group linkages than by the content of the proposal (Galligan 1990, 2001). A referendum which involves a cleavage or ideological issue, and/or in which political parties take well known and predictably opposite positions, ought to hold the least potential for opinion change. One which involves a new or previously undiscussed issue, or in which parties line up in a non-traditional manner, is more likely to promote some of the short-term variables towards the right side of figure 2. But other dynamics can sometimes occur. The 1986 Irish divorce referendum might have seen less movement over the course of the campaign had it been fought solely along religious or partisan lines. But the rather dramatic shift which took place in voter sentiment during that campaign was attributable in part to the success of campaign actors in persuading voters to view the matter as something other than a traditional cleavage issue (Darcy and Laver, 1990). Similarly, many Australian voters became persuaded over the course of the 1999 referendum campaign to view the choice in terms of an elected or appointed presidency rather than one of maintaining or abolishing the monarchy. While attitudes toward the latter question might also conceivably have changed over the course of a campaign, they almost certainly would have been more stable in the aggregate than those involving the newer and (up to that point) less widely discussed issue of an elected Australian presidency (Higley and McAllister, 2002). Thus, an important part of the discourse of a referendum campaign often involves one side or the other attempting to change or redefine the subject matter. Europe in some sense is undoubtedly well on the way to becoming a cleavage issue in British politics (Evans, 1999). But how strong a cleavage? And for how many voters? Also, to the extent that it cross-cuts party lines, it may not be the defining cleavage in a referendum, as NATO was not for Spanish voters in spite of the strength of opinion on the issue (Boix and Alt, 1991). I will return to this theme at a later point in the paper. But it should be noted here that, even when a referendum involves an issue on which there are strong preexisting cleavages, the nature of the campaign may be such that the existing division of opinion on the underlying issue may not be the key factor in determining the outcome. It may seem perverse to argue that a referendum on the Euro or on an EU constitution need not necessarily be about Europe. The way in which the issue is framed, together with the dynamic of the campaign, is equally likely to determine what voters believe they are voting on. There are numerous examples in the comparative literature on referendums showing that the outcome of a referendum can be quite different than the division of public opinion on the underlying issue, sometimes even when such opinions are strongly held. 5

6 The electorate Turnout tends to fluctuate more widely in referendums than it does in national elections. Butler and Ranney (1994) found that turnout over a large number of referendum cases in various nations averaged fifteen percentage points lower than that found in general elections in the same countries. It can however sometimes rise to much higher levels when a particular issue engages wide voter interest or when a more intense campaign is waged by interested groups. The turnout in some of the more important European referendums has often been higher than that found in the most nearly comparable elections in the same countries (table 2). Turnout in the Quebec sovereignty referendums was likewise higher, and voter participation in the 1995 referendum registered an astonishing 94%, a full twelve percentage points above that of the provincial election held a year earlier. Other important referendums in which turnout registered higher than that of a comparable election (table 2) are the 1992 Canadian constitutional referendum (+5), the 1994 Norwegian EU membership referendum (+13) and the Danish and Swedish referendums on the Euro (+1, +3). The record in Britain to date is mixed, as turnout in the 1975 referendum or in the 1997 Scottish and Welsh referendums was somewhat lower than in a comparable general election. And of course turnout in British elections has been falling more generally. But even at these levels, the question of mobilization becomes important. When turnout is low, the ability to mobilize one s own supporters counts for more. When it rises, it is generally because the issue itself is perceived as an important one for most voters, sometimes generating new sources of participation. Given the Danish and Swedish experiences, there is some reason to believe that turnout in a British referendum on the Euro might be at least comparable to that of a general election. But, in a referendum on the constitution, there is a chance that it could be much lower, perhaps closer to that of European parliament elections. At those levels, mobilization of partisan voters could count for a great deal, whether they be Labour or UK Independence supporters, as in the most recent European election. The electorate found in a referendum is not always the same one that participates in elections. However, the same factors that predict participation in an election(e.g. age, education, income, political interest) generally tend to correlate with referendum voting (Jenssen et al, 1998). Saint-Germain and Grenier (1994) found a strong resemblance between patterns of vote for the Parti québécois in the 1989 Quebec provincial election, the vote for the Bloc québécois in the 1993 federal election, and the NO vote in the 1992 Canadian constitutional referendum -- with turnout in all three instances fluctuating within a fairly narrow range. The wider variations in turnout found in referendums partly explain their greater potential for volatility. They also suggest that turnout itself is at least partially a campaign effect. Where parties fail to mobilize their supporters on behalf of an issue, or where non-party groups succeed in mobilizing theirs, the outcome of a referendum can be more directly subject to a differential turnout effect. Since the turnout in some referendums can actually be higher than that found in elections, even though on average it is lower, the potential for turnout effects in a referendum is generally greater than that found in ordinary parliamentary elections. In a British referendum on the Euro, and/or on an EU constitution, turnout is likely to be an important factor in determining the outcome. 6

7 Table 2. Selected turnout comparisons referendums and elections % +/- Austria 1994 EU membership referendum 81-1 " 1994 parliamentary election 82 Canada 1992 constitutional referendum 75 5 " 1993 federal election 70 (Quebec) 1980 sovereignty-association referendum " 1981 provincial election 83 " 1994 provincial election 82 " 1995 sovereignty referendum Denmark 1992 Maastricht treaty referendum 83 0 " 1993 Edinburgh agreement " 1994 general election 83 " 2000 European currency referendum " 2001 general election 87 Finland 1994 EU membership referendum 71-1 " 1994 presidential election (2nd round) 82 " 1995 parliamentary election 72 France 1992 Maastricht treaty referendum " 1993 national assembly election 69 " 1995 presidential election (2nd round) 75 Norway 1993 general election 76 " 1994 EU membership referendum Sweden 1994 EU membership referendum 83-4 " 1994 general election 87 " 2003 European currency referendum " 2002 general election 80 UK 1975 EC membership referendum (October) general election general election European parliament 39 (Scotland) 1997 devolution referendum " 1997 UK election (Scotland) Scottish parliament 59 (Wales) 1997 devolution referendum " 1997 UK election (Wales) 73 " 1999 assembly 46

8 Referendum campaigns In a referendum, campaigns are frequently more important than in ordinary partisan or candidate elections. In some referendum campaigns, public opinion has sometimes been seen to shift dramatically, even in some instances where the distribution of opinion on the issue of the referendum was well known and seemingly well established. In several of the referendum cases mentioned in this paper, polls taken early in the campaign period would have suggested quite different results from those which actually occurred. The dynamics of a referendum campaign can be harder to anticipate than those of an election, and the participation of the electorate varies more widely. Context is all important. The political and economic circumstances in which the vote takes place, the images that voters hold of the groups and individuals involved, and their reactions to the specific discourse of the campaign, can be as important to the voting decision as opinions on the actual ballot question. While longer term factors such as partisanship or ideology can also be of considerable importance to the voting decision, the short-term impact of campaign strategies and tactics are often critical factors in determining the outcome, especially in those instances where the issue(s) of the referendum or initiative are new to the voter. In an earlier paper (LeDuc, 2002a), I conducted an analysis of the magnitude of opinion change in a number of referendum campaigns by comparing their outcomes with a result which would have been predicted by a public opinion poll taken one or two months before voting day. 1 In this way, a rough aggregate measure of the magnitude of campaign effects, somewhat similar to a Pedersen index, was obtained. 2 That exercise demonstrated that, in 23 referendum cases for which suitable poll data were available, the average movement over the course of the campaign was 17 percentage points, a figure substantially higher than that obtained by comparing opinion or vote intention change over an election campaign period for the same set of countries. In most of these instances, the direction of the movement was negative i.e. public support for a ballot proposal deteriorated over the course of the campaign. Thus, a referendum might be initiated by a party or government based on a particular set of polling assumptions, only to discover that its strategic advantage suddenly vanishes in a short, intensive campaign. A summary of the patterns of campaign movement for some of the cases which might be relevant to future British referendums on European questions is shown in figure 2. These cases represent a considerable range of variation in countries, issues, and political context. In most of these instances, the magnitude of the shift in public opinion occurring over the course of the campaign period was more than enough to determine whether a referendum on a particular issue succeeded or failed. It cannot escape notice that the direction of movement is negative in many of these cases, in some instances largely negative. The type of referendum campaign which tends to display the greatest campaign volatility is frequently one in which there is little partisan, issue, or ideological basis from which voters might tend to form an opinion. Voters cannot generally be expected to have well formed opinions on an issue that has not previously been a subject of any broad public debate. Some referendums fitting such a profile are those that involve multiple issues, complex international treaties, or large packages of constitutional provisions. The 1992 Canadian constitutional referendum and some Australian referendums display this pattern, with 8

9 Figure 2. Net opinion change in seventeen referendum campaigns 1992 Maastricht Denmark France Ireland Quebec Devolution Scotland Wales Ireland divorce EU Membership Austria Finland Norway Sweden Canada 1992 Australia 1999 European currency Denmark 2000 Sweden 2003

10 elite driven projects being decisively rejected once the voters had learned enough about them (LeDuc 2003; LeDuc & Pammett 1995; Johnston et al 1996; Galligan 1990, 2001). The 1992 referendums on the Maastricht treaty, particularly in France, displayed many of these same characteristics. In such circumstances, the degree of change in opinion over the course of even a short campaign is potentially large, because there is little in the way of stable social or political attitudes, or partisan cues, which might anchor opinions on the issue. Referendums on issues which have been debated extensively in political arenas other than that of the referendum campaign, or in which there were strong linkages to the positions taken by political parties, generally displayed less campaign volatility.. Here, opinion tended to be much firmer and less subject to rapid change or sudden reversal. The voting decision in such circumstances is often easier for many voters, and tends to be made earlier in the campaign. There are often strong cues based on partisanship or ideology, and campaign arguments are advanced by familiar party leaders. In such a campaign, much of the attention is directed toward wavering or undecided voters, in the knowledge that a swing of only a few percentage points might make the crucial difference in the outcome. The 1995 Quebec sovereignty referendum is a good example of this type of referendum campaign. Because this issue was by that time such a familiar one to Quebec voters, and because the parties were so highly mobilized on the core issue, the degree of movement in public opinion caused by the campaign was much less. So too in the 1994 referendums on EU membership in Austria and the Nordic countries, since EU membership had been actively debated for a considerable period of time prior to the referendums. A similar pattern may have occurred in the case of the referendums on the European currency in Denmark and Sweden, because European issues, and particularly the Euro, have likewise been widely debated in recent years, and electorates in both countries remain sharply divided on many of the larger questions of European integration. The nature of the issue appearing on the ballot, and the political context in which the referendum takes place, nevertheless explains only part of this variation in the level of campaign effects in different types of referendum cases. In some instances, a entirely unexpected campaign dynamic can develop when a referendum on a reasonably well known issue begins to take on a new and unanticipated direction over the course of the campaign. Sometimes, this occurs when opposition groups are successful in changing the subject of a referendum, or in raising doubts about the motives of those who proposed it. Darcy & Laver (1990) documented this type of campaign in their study of the 1986 Irish divorce referendum, describing the dynamic in that instance as one of opinion reversal. A similar dynamic may have occurred in the case of the 1999 Australian referendum on the monarchy, in which the campaign rhetoric persuaded many voters to view the choice as one between an elected and appointed presidency rather than between retaining or abolishing the constitutional role of the British monarch in Australia (Uhr 2000; Higley & McAllister 2002). In both the 2000 Danish referendums and the 2003 vote in Sweden, proponents of the Euro tried to keep the campaign debate narrowly focussed on technical and monetary issues. But opponents in both instances continually shifted the debate toward larger issues, attempting to convince voters that the issues at stake in the referendum were much greater. 10

11 The referendums on the Euro held in Denmark in 2000 and in Sweden just a year ago provide an illustration, and to some degree a test, of some of the above arguments about the characteristics of referendum campaigns. Since Denmark first joined the European Community in 1973 (at the same time as the UK), the European issue has been at the center of Danish politics. Over the past thirty years, Denmark has held six referendums on various European issues, including the 1973 decision to join. 3 Although the vote against the Single European Act in 1986 was substantial (44%), the watershed event of this period was Denmark s narrow rejection of the Maastricht Treaty in the June 1992 referendum. This was an event that sent shock waves throughout Europe, as it represented a major political setback at the time for the next stage of the European project. Certainly, the Danish vote, along with the issues that it raised, was a factor in the referendum on Maastricht held in France three months later. Although a compromise which allowed several specific Danish exceptions to Maastricht was cobbled together in Edinburgh and ratified in another referendum barely a year later, the Danish vote against Maastricht had lasting effects. Out of the Danish NO campaign in the 1992 referendum was born one of the first well organized political groups (the June Movement) that articulated the reservations than many outside the political elites felt about the speed and direction of European integration. From 1992 onward, Euroskepticism as it came to be known in Britain, became a powerful political force in a number of EU countries. But it was in Denmark that it first found its political voice. Figure 3. Vote intention in the Danish EMU referendum, March-September 2000 * denotes actual result

12 Given this background, it was not surprising that the prospective entry of Denmark into the European Monetary Union (EMU) proved to be a politically contentious proposition. It was clear from the beginning that there would have to be a referendum on the issue, in spite of the broad parliamentary support, and that the chances of passage in such a referendum were doubtful. Nevertheless, the outcome of the vote conceivably might have gone either way. Prime Minister Poul Nyrup Rasmussen led the YES campaign ( Best for Denmark ), and utilized the considerable resources of the government on its behalf. The timetable allowed the government to plan its campaign well in advance and to anticipate many of the sources of opposition. The date of the referendum was announced in March, at a time when the public opinion polls appeared favourable., with some showing a YES lead of as much as 15%. All of the mainstream political parties, comprising 140 of the 179 seats in the Folketing supported Danish participation in the EMU. 4 Also lined up in support of the Euro were the leaders of the major trade unions, nearly all elements of the business community, and most of the nation s newspapers. On the surface, it appeared that the chances of success in such a setting were fairly good. (Downs, 2001). But by mid summer, the YES lead had evaporated. Polls in June showed the race at a dead heat, and the number of undecided voters in the percent range (figure 3). The government s principal strategy was to try to convince voters that entry into the Eurozone was a necessary step to protect the Danish economy. A considerable part of the YES campaign therefore stressed the potential economic dangers of continued Danish nonparticipation in EMU slower economic growth, a weaker currency, higher interest rates and unemployment, etc. The YES side also sought to separate the Euro question from larger European issues, stressing that EMU was an economic issue that did not necessarily entail closer political union. This strategy however was undermined by a report published by the Danish Economic Council (dubbed the three wise men ) in May. This report stated quite clearly that the Euro would make little difference to the Danish economy, since the krone was already pegged to the Euro. In the view of the economists, the issue was a purely political one. Although the report said little that was really new, its timing and conclusions proved highly damaging to the YES campaign (Qvortrup, 2001). The Danish NO campaign in contrast was more diverse and wide ranging. It did not so much have to make a coherent case against the Euro as to raise doubts and suspicions about the arguments being put forward by the YES side. Ranging from far-left to far-right, NO campaigners had various reasons for their opposition to Danish entry to EMU. Some such as Pia Kjaersgaard, the leader of the small Danish Peoples Party, drew heavily on patriotic and nationalistic themes. Others raised the spectre of a powerful European central bank, or portrayed EU institutions and practices as a threat to Danish democracy. The left parties that opposed the Euro tended to stress potential threats to the safeguards of the Danish welfare state. The NO campaign was also aided inadvertently by the Euro itself, which had been declining steadily against other currencies, particularly the U.S. dollar, since its official launch in January By voting day, the outcome appeared certain, although the YES side seemed genuinely shocked by its loss. 12

13 Figure 4. Vote intention in the Swedish EMU referendum, January-September 2003 * denotes actual result Except for the dramatic and unanticipated events which occurred in the final week of the campaign, the Swedish referendum exhibits some similarity to that the one in Denmark on the same issue three years earlier. In Sweden, the narrow vote in favour of joining the EU in 1994 (52% YES) revealed deep divisions over the issue, both within the governing Social Democratic Party, and in the country more generally. Sweden s decision to remain outside the Eurozone reflected the government s cautious handling of the issue in the wake of the narrow1994 vote. In both Denmark and Sweden however, it was realized that a vote on the Euro would have to take place in a matter of time. In both cases, it would be the governing party that would determine the timing, and to a certain extent the context, in which such a vote would occur. Of course, context as discussed here can be determined only in a limited way by the political leaders who set the terms of the referendum. While they can determine the date of the vote, the wording of the question, and to some degree the connection of the issue to other elements of public policy, they cannot change the history of the issue, or manipulate the predispositions of much of the electorate on the larger question of European integration. Nor can they control many of the forces of the campaign, once it is under way. In other words, the dynamic of a short intensive campaign on an emotionally charged issue can only partly be anticipated in advance. 13

14 Of course, Swedish politicians must have had the Danish example firmly in mind when they proceeded to call the referendum on the Euro. In Sweden, there was even less reason for supporters of EMU to be optimistic about the prospects for its passage, although some polls taken early in 2003 suggested the possibility of a close contest, perhaps similar to that on EU membership in Over the course of what might be called the long campaign, the spread favouring the NO side steadily widened (figure 4). However, the percentage of undecided respondents found in the polls was invariably high, sometimes as much as a fifth or a quarter of the potential electorate. In spite of the seeming odds against them, supporters of the Euro appeared optimistic that these wavering voters could be won over by a strong campaign. The decision to place the foreign minister, Anna Lindh, at the forefront of the YES campaign was a strategic one, allowing the government s position to be articulated by its most popular political figure and shifting attention away from the prime minister and other policy matters. It had also been hoped that Lindh s role in the campaign might help to close the gender gap in public opinion, as polls in Sweden had consistently shown a higher percentage of women favouring the NO side. Her tragic murder four days before polling day threw the campaign into disarray, and prompted much speculation about the possibility of a sympathy vote, which might tip the balance to the YES. Indeed, the very last polls taken just before the vote appeared to show a narrowing of the spread between the two sides. Although Lindh s assassination effectively suspended campaigning in the final days, it is likely that the result was determined by other elements much earlier in the campaign. As had been the case in the Danish referendum three years earlier, the government found it difficult to make the case for the Euro on narrow economic grounds without opening up the larger political questions of European integration. The background debate taking place on a proposed European constitution made this segregation of issues all the more difficult. The fact that the Swedish economy was strong, while that of Germany was in recession, further undermined many of the arguments that were advanced about potential risks to the economy in rejecting the Euro. Also, as had been the case in the 1994 referendum on EU membership in Sweden, splits within the cabinet itself opened up with some frequency over the course of the campaign, causing further damage to the YES side. The Swedish referendum, in spite of the unique and tragic events associated with it, conforms to all of the theoretical arguments advanced here regarding the dynamics of referendum campaigns. There can be little doubt that levels of predisposition on the issue were high, given the history of the European issue in Sweden and the other Nordic countries. But this does not mean that the campaign was irrelevant. Such predispositions invariably interact with information provided by the campaign and its actors. In this case, the credibility of the government s message appears to have been an important element, as it was also in the Danish referendum on the same issue. Perceptions of the state of the economy also played a role, but in Sweden it was a strong economy that had the effect of undermining the government s strategy rather than enhancing it. A dramatic campaign event had at least the potential to produce extensive disruption, although in the end it appeared not to do so. And, as had been the case in 1994, divisions within the governing party also played a role, undermining the link between opinion and partisanship for at least some voters. 14

15 One important difference between the Danish referendum on the Euro in 2000 and the Swedish referendum of 2003 is that, in Denmark, the right wing political parties, principally the Danish Peoples Party, and the organized movement of Euroskeptics (the June Movement) which had campaigned effectively against Maastricht, provided much of the leadership of the NO campaign. In Sweden, opposition to the Euro was centered mainly on the left of the political spectrum, and some of it was found within the governing Social Democratic Party itself. Thus, the themes and issues of the NO campaign were different in the two countries. In Denmark, the NO campaign often invoked patriotic or nationalistic themes, and stressed the potential loss of Danish sovereignty in a more integrated Europe. In Sweden, the themes of the NO campaign centered around the structure of the Swedish welfare state, and the threats to that structure which might be posed by powerful European institutions with weaker commitments in the area of social policy. In both cases, the arguments advanced had little to do with the Euro per se. But they indicate the difficulty that a governing party faces in attempting to structure such a campaign around a specific issue. Also, they demonstrate the innate advantage often held by the NO side in referendum campaigns, and perhaps explain in part why the short-term trend in public opinion in such campaigns is so often towards the negative side. 5 NO campaigners do not necessarily have to make a coherent or persuasive case against a proposal. Risk aversion can be a powerful basis on which to appeal to some voters in and of itself, and the NO side of a referendum campaign usually (not always) holds a monopoly on this emotion. The shift in public sentiment which frequently takes place in a short campaign need not be permanent. Sometimes, public opinion on a widely debated issue can be seen to shift back again, once the turbulence of the referendum campaign is out of the way. However, in the Swedish case, given the long history of the European issue, the levels of predisposition, and the relatively small movement of the polls over the course of the long campaign, it seems unlikely that the YES side could have prevailed in any event. The vote decision We would expect that in those instances where the issues of the referendum are entirely new to the voter, the learning process of the campaign should be more critical for deciding how to vote and therefore also more important in determining the outcome. In those cases where voters clearly need the campaign in order to form an opinion on the issue(s) of the referendum, we would expect more individual voting decisions to be made late in the campaign, after a sufficient amount of information has become available about the issue on which voters are being asked to render a decision. Conversely, where voters are able to make up their minds on the basis of clear partisan or ideological cues, or where there is a high degree of prior familiarity with the issue(s) of the referendum, we might expect voting decisions to be made earlier. The timing of the vote decision therefore may be a useful indicator of the extent to which the characteristics of the referendum vote decision vary from that typically found in an election. Survey data on reported time of vote decision are available for several of the referendums mentioned here, although not for the recent Danish and Swedish cases. Table 3 15

16 presents data for ten referendums, a number sufficient to provide some variation in context in terms of the amount of prior knowledge that a voter might have been expected to have regarding the issue being voted upon in the referendum. The 1992 Canadian constitutional referendum provides the most extreme example, because the referendum was not anticipated and voters could not have been expected to have a high degree of prior knowledge of the content of a constitutional agreement which had been negotiated in closed sessions. Not surprisingly therefore, nearly two-thirds of those voting in that referendum made their decisions over the course of the campaign, a substantial number of these as late as the final week. By contrast, voters in the 1995 Quebec sovereignty referendum were able to come to a decision much more quickly on the issue, in part because the subject matter of the referendum was so well known, but also because the campaign provided strongly reinforcing partisan cues for many voters. While the campaign was still important to the outcome, in part because of the closeness of the result, fewer voters needed the additional information provided by the campaign in order to reach a decision. Three-quarters of the Quebec electorate had already made up their minds how to vote at the time that the referendum was called. The 1994 European Union membership referendums in the Nordic countries, provide examples which fall between these two extremes, as does the 1992 French referendum on the Maastricht treaty. In these four cases, more than half of all voters surveyed reported having made their decision how to vote long before the campaign had begun. The balance decided how they would vote at some point over the duration of the campaign, dividing roughly evenly between those reporting that they made their decision in the final week and those deciding earlier in the campaign. It is logical that these cases would fall in the middle of the distribution suggested by table 3, because voters in the EU referendums would have had a high degree of knowledge of the underlying issue, but would still have needed the campaign discourse in order to assess the specific arguments regarding the accession agreements which were being put forward by the parties, or, in the case of France, the actual content of the Maastricht treaty and the interpretations being placed on it by those involved in the campaign. The fact that parties that are normally opponents in election campaigns were campaigning together in support of EU membership in several of the Nordic countries also have served to present voters with new information, in which it could be expected that more time might be required for this to be factored into the decision. In Sweden, divisions among the governing Social Democrats spilled over into the 1994 campaign, with the government actively supporting the YES side but some of its own partisans campaigning against it under the umbrella group Social Democrats Against the EU. Similar divisions within the governing party were apparent in the 2003 referendum on the Euro. These circumstances present a quite different picture than the 1995 Quebec case, in which parties with well known and strongly held positions on the sovereignty issue were putting forward highly familiar arguments right from the beginning and mobilizing their traditional supporters in support of a cause to which they could be expected to have a strong prior commitment. 16

17 Table 3. Reported time of vote decision in ten referendums (%) Long before At call During campaign Final week Quebec Finland France Norway Sweden Quebec Australia Scotland Wales Canada Carleton ISSP Study 2 Comparative Nordic Referendums Study 3 SOFRES/ Le Figaro (Franklin, van der Eijk and Marsh, 1995) Canadian National Election Study: Quebec referendum wave Australian Constitutional Referendum Study CREST surveys Carleton Referendum Study The four remaining cases for which data on time of vote decision are presented in table 3 the 1980 Quebec sovereignty referendum, the 1999 Australian republic referendum, and the 1997 Scottish and Welsh devolution referendums are ones in which a somewhat larger share of the electorate is found deciding later in the campaign. In these cases, like that of the 1992 Canadian constitutional referendum, the actual call of the vote was important, because it provided the first real information to voters about the decision to come. The announcement of the Scottish and Welsh votes came only after Labour s victory in the May 1997 election, and Blair s subsequent decision to make these reforms his first legislative priority. In the 1980 Quebec case, the forthcoming sovereignty referendum had been the subject of discussion for some time, in part because the PQ had made a commitment to it during its 1976 election campaign. But no one knew exactly when the referendum would take place or what the wording of the question would finally be until the date of the referendum was announced. In Australia, a referendum on the monarchy had also been the subject of political debate since the former Labor prime minister, Paul Keating, signalled his government s intention to propose a constitutional amendment to end the role of the British monarch as the Australian head of state. But Labor s defeat in the 1996 election shifted the course of that debate and the new Liberal government 17

18 referred the matter to a convention. The question ultimately put forward to the electorate by that convention was significantly different than the original proposal and included a controversial preamble on which voters were also asked to render a decision. 6 This continual flow of new information, both on the proposal itself and on the nature of the debate, would have made it difficult for all but the most strongly committed republicans among Australian voters to have made up their minds very far in advance of the referendum. It is of course uncertain how much advance notice British voters may receive regarding a referendum on the Euro or on a new EU constitution. An announcement of an agreement on the new constitution, followed by a quick vote, could produce a scenario somewhat like the 1992 Maastricht campaign in France, in which disagreements about the meaning of specific provisions of the constitution form a substantial part of the campaign discourse. Such campaigns are generally more volatile, because of the continual flow of new information to voters from competing sources. There could also be a domino effect, if several countries hold referendums on the new constitution within a short space of time (Jahn and Storsved, 1995). A referendum on the Euro, however, would be a solely British event. Particularly if announced well in advance, it is more likely to follow a dynamic somewhat like that of the 1994 EU referendums, in which larger numbers of voters are able to make up their minds earlier. The linkage between pre-existing partisan attitudes and the referendum issue can tell us much about the influences of predispositions and the flow of information derived from the campaign. Ideology, where it forms a substantial underlying component of the party system, can similarly predispose a voter to a particular position on a referendum issue. The positions which political parties take, either in the evolution of the referendum issue or during the campaign itself, provide one of the strongest available information cues to voters. Where these are present, voters with partisan predispositions are able to find their own positions on an issue fairly quickly. Where they are absent, other short-term elements tend to become more powerful, or well-known groups, organizations, or individuals may intervene to generate the same types of short cuts that political parties might otherwise provide (Lupia and McCubbins, 1998; Bowler and Donovan, 1998, 2002). In the following examples, I draw upon surveys which contain data on partisanship, ideology, attitudes toward the referendum issue (or its underlying components), and voting choice (or intention) in order to measure the relative effects of partisan or other predispositions on voting choice in several past European referendums. Table 4 displays the relationships found in the Nordic countries between feelings toward the parties and feelings towards the EU, as well as the correlations between party feelings and voting behavior in the 1994 referendums on EU membership. 7 In Norway, for example, the correlation between feelings about the EU and feelings toward the Labour and Conservative parties, both of which campaigned for EU membership, are moderately strong and positive (.32 and.39 respectively). Yet the Conservatives managed to deliver a much higher percentage of their supporters to the YES side in the referendum (85%) than did the governing Labour party (65%). Supporters of the Centre party, which campaigned vigorously against membership, voted almost entirely NO. This is not surprising when one considers the strong correlation between 18